How the Electric Grid Works

Key Takeaways

• The US has three largely separate grids (Eastern, Western, Texas)
• RTOs manage wholesale markets and coordinate reliability across states
• Transmission (high-voltage, long-distance) differs from distribution (local delivery)
• Behind-the-meter generation bypasses queue delays but creates other challenges

Three Nations of Electricity

When most Americans think about the electric grid, they imagine a unified national network seamlessly delivering power from coast to coast. The reality is far more fragmented. The United States doesn't have one grid—it has three largely separate electrical networks that operate with minimal connection to each other.

The Eastern Interconnection stretches from the Atlantic Ocean to the Great Plains, serving roughly two-thirds of the US population. The Western Interconnection covers everything from the Rockies to the Pacific, plus parts of Canada and Mexico's Baja California. And then there's Texas.

ERCOT—the Electric Reliability Council of Texas—operates the Texas Interconnection, an isolated grid serving 90% of the state's load. This isolation is deliberate: by keeping the grid almost entirely within state borders, Texas avoided federal jurisdiction under the Federal Power Act. What started as a regulatory maneuver in the 1930s has shaped everything from winter storm vulnerability to data center development strategy eighty years later.

These three grids connect through a handful of high-voltage DC interties, but the transfer capacity is minimal—typically measured in hundreds of megawatts when the grids themselves operate at hundreds of gigawatts. When Winter Storm Uri knocked out 52 GW of Texas generation in February 2021, the Eastern and Western grids could do almost nothing to help. Fragmentation matters.

The Structure of the Grid

The path from power plant to your phone charger involves three distinct layers, each operating at different voltage levels and serving different functions.

Generation produces electricity at power plants—coal, natural gas, nuclear, wind, solar, hydroelectric. Most generators produce power at 10-25 kilovolts (kV).

Transmission moves that power long distances at high voltages—typically 115 kV, 345 kV, or 765 kV. Higher voltages reduce energy loss over distance (resistive losses scale with the square of current, and voltage determines current for a given power level). The transmission network consists of high-voltage lines, towers, and substations that form the interstate highway system of electricity.

Distribution delivers power locally at progressively lower voltages—first to neighborhood substations at 12-35 kV, then to homes and businesses at 120-240 volts. This is the network of wooden poles and underground cables you see in residential areas.

Understanding this structure is critical for data centers. A 200-home neighborhood might require 2 megawatts from the distribution network. A gigawatt data center requires 500 times that capacity. Distribution infrastructure simply cannot handle such loads—these facilities must connect directly to the transmission system, often requiring construction of dedicated substations and miles of new high-voltage lines.

RTOs: The Traffic Controllers

In most of the country, Regional Transmission Organizations (RTOs) or Independent System Operators (ISOs) manage the wholesale electricity market and coordinate grid operations across utility territories. Think of them as air traffic controllers for electrons.

PJM Interconnection is the largest RTO, covering 13 states and the District of Columbia—from New Jersey to Illinois, serving 65 million people and coordinating a $50 billion annual wholesale market. On any given day, PJM's computers process thousands of generator bids, forecast demand, calculate optimal dispatch, and ensure voltage and frequency remain stable across hundreds of thousands of square miles.

Other major RTOs include MISO (Midcontinent Independent System Operator) spanning 15 states from Montana to Louisiana, SPP (Southwest Power Pool) covering the Great Plains, and CAISO (California ISO) managing the California grid.

When you're trying to connect a gigawatt data center, the RTO jurisdiction matters enormously. Different RTOs have different interconnection procedures, different queue backlogs, different cost allocation methodologies, and different approval timelines. A project in PJM faces a fundamentally different regulatory path than one in ERCOT or MISO.

Data Center Alley: A Case Study

Northern Virginia has become the densest concentration of data center capacity in the world—over 6 gigawatts and growing. But this success has created its own constraints, offering a preview of challenges spreading nationwide.

The region's dominance began with AOL in the 1990s, accelerated with the growth of Amazon Web Services in the 2000s, and exploded as hyperscalers consolidated operations in areas with favorable connectivity, policy, and power. Loudoun County alone hosts over 300 data centers, consuming more electricity than many entire states.

But Dominion Energy, the utility serving the region, faces a fundamental physics problem: you cannot add infinite load to a fixed transmission infrastructure. The utility is now building billions of dollars in new transmission capacity, installing new substations, and extending timelines for new connections. What once took 18-24 months now takes 4-6 years.

This is why data center operators are looking beyond Virginia—not because the incentives have changed or fiber is unavailable, but because the grid cannot absorb the load. Physical infrastructure, not policy or economics, has become the binding constraint.

The Interconnection Process

Connecting a large load to the grid isn't like plugging in an appliance. It requires years of engineering studies, cost assessments, and infrastructure construction coordinated across multiple utilities and the RTO.

The traditional process involved three sequential studies:

1. Feasibility Study - Does the grid have capacity in this location? What would connection require at a high level?
2. System Impact Study - How would this load affect grid stability? What upgrades are needed to maintain reliability?
3. Facilities Study - Detailed engineering and cost estimates for required upgrades.

Each study could take 6-12 months. Projects moved sequentially through the queue—first come, first studied. But the queue backlog exploded: PJM alone has seen over 270 GW of requests (compared to its total installed capacity of ~185 GW). Some projects wait 5-8 years just to complete studies, never mind actual construction.

The bottleneck became so severe that FERC (Federal Energy Regulatory Commission) issued Order 2023 in 2023, mandating a shift to cluster studies—batching projects together to study grid impacts simultaneously. In theory, this speeds up the process. In practice, utilities still face engineering capacity constraints, and the backlog remains enormous.

This is why site selection must begin with transmission topology. You can negotiate land, permits, and incentives in months. Grid connection approval takes years—and you only discover the full cost after completing studies. A site that looks perfect may become infeasible when the utility reveals a $200 million transformer upgrade requirement buried in the facilities study.

ERCOT: Faster but Riskier

Texas operates under different rules. Because ERCOT doesn't cross state lines, it escapes FERC jurisdiction for most operational matters. The state's Public Utility Commission of Texas (PUCT) oversees interconnection, and the process moves faster than in RTO territories.

Where PJM might take 5-8 years from application to energization, ERCOT can complete the same process in 2-4 years. The utility files a Certificate of Convenience and Necessity (CCN) with the PUCT, constructs the required facilities, and the interconnection proceeds. Fewer jurisdictions, fewer handoffs, less bureaucratic overhead.

But speed comes with trade-offs. ERCOT's isolation meant it couldn't import power during Winter Storm Uri. The grid came within minutes of complete collapse—a scenario that would have required weeks or months to restore and caused hundreds of billions in economic damage. Over 200 people died in the immediate aftermath, with estimates ranging higher when accounting for indirect effects.

Despite these risks, data center operators continue choosing Texas for projects that prioritize speed to market. The calculus is straightforward: accept isolation risk in exchange for cutting years off the development timeline. For AI inference capacity needed in 2026, an interconnection approval delivered in 2028 is worthless.

Behind-the-Meter: Bypassing the Queue

If the interconnection queue creates a 5-8 year bottleneck, why not simply generate power on-site?

Behind-the-meter generation means producing electricity at or near the point of consumption, bypassing the interconnection queue entirely. A data center might install natural gas turbines, build a solar array with battery storage, or negotiate a dedicated connection to a nearby power plant.

The advantages are clear:

• Speed - No queue waiting, faster project approval
• Control - Direct management of power supply and quality
• Redundancy - On-site generation provides backup during grid outages

But behind-the-meter comes with significant disadvantages:

• Cost - On-site generation typically costs 2-3x grid power on a levelized basis
• Emissions - Natural gas turbines produce direct CO2, complicating sustainability commitments
• Fuel logistics - Natural gas pipelines require their own approval process; diesel requires trucking
• Maintenance - Operating a power plant is a completely different business than operating data centers

Some operators pursue hybrid approaches: behind-the-meter for initial capacity while waiting for grid interconnection, then transitioning to grid power for base load. Others are exploring partnerships with nuclear plants or dedicated renewable installations.

But behind-the-meter remains a workaround, not a solution. At scale, it's simply more expensive and operationally complex than grid power. The fact that operators seriously consider it reflects how severe the grid bottleneck has become.